TREE vol. 3, no. 8, August 1988

Lennart Hansson and Heikki Henttonen

quantity and quality of food rangefrom simple hypotheses ofovergrazing4 to ideas about vari-ation in plant production5, nutrientcycling6, hormonal precursors7 and,more recently, plant defence8.Same ideas about the significanceof habitat heterogeneity andpatchiness may also be related tofood resources.

Most rodent species have spe-cialized diets and feedingstrategies, e.g. granivory, folivoryand bryovory, .and have differentamounts of food available to them.For multispecies rodent communi-ties in which there are several feed-ing habits, aspects of the biology ofa particular food-plant group do notexplain synchronous populationpattems in the 'rodent species, butmay be sufficient to explain non-synchronous anes. For example,hormonal precursors found ingraminids obviously cannot explaina synchronous decline in shrub andmoss eaters, though they mayaccount for the seasonal breedingpattem of grass eaters. If rodentsare limited only by food, then ex-periments with supplemental foodshould have clear effects onpopulation dynamics.

Plant production cycles ('fIower-ing cycles') have been thought toresult from an interaction of plantgrowth rhythms and temperaturesin harsh climates5. The plant cycleshould be independent of grazingeffects. As the response to temper-ature differs between plan t spe-cies, the dynamics of the consumerspecies should also differ, even ifwarm summers tend to synchronizethe plant production. Nutrient cyc-ling due to small rodents them-selves could affect other species inthe same habitat. Nevertheless, fol-ivores and granivores, eating plantparts with different nutrient con-tents, should react differently fromeach other.

Habitat heterogeneity can affectthe amount of food available. If theamount of suitable habitat for aspecies is small in winter (or atother seasons) in relation to itstotal area, or if there are largeseasonal differences in habitat

Lennart Hansson is at the Department of Wildlife Extrinsic factorsEcology, Swedish University of Agricultural Sd- Of the possible extrinsic factorsences, P.O. Box 7002, 5-750 07, Uppsala, Sweden affecting population dynamics,and Heikki Henttonen is at the Forest Research food has received the most atten-Institute, P.O. Box 18, SF.Q1301 Vantaa, Finland. tjon. Suggestions for the role of the

@ 1988. Elsevier Publications. Cambridge 0169-5347/88/$0200 195

many textbooks as examples of regularpopulation cyc/es with constant intervaland amplitude. However, recent evidenceand analyses have indicated much morecomplex patterns, with geographic trendsin frequency and amplitude of ffuctuationsand covariation wlth many interactingcommunity components. These newfindings indicate that extrinsic factors aremuch more important for the generation ofregular rodent cyc/es than was earlierbelieved, and that regular cyc/es representonly a minority of the dynamic patternsfound in rodents.

Ideas about general and regulardensity variations in rodents goback to reviews from the 19605 and197051. The popular view at thattime was that this supposed regu-larity was due to intrinsic, sodaland/or genetic factors. However, atthe same time a few authors stillthought in terms of extrinsic factors,such as disease, food and pred-ators, as driving processes ofpopulation cycles.

The conceptual. basis for under-standing rodent dynamics haschanged considerably during the19805, and evidence has accumu-lated showing that the key par-

°ameters of population fluctuations(amplitude, frequency and inter-specific synchrony) vary geographi-cally. Here we review the new data. and examine the extent to which

geographical trends in rodentdynamics are correlated withphysical and biological variables.Such correlations may be due torodents affecting other species in aunidirectional way (e.g. prey topredator). Alternatively, species inother trophic levels may dedde thegrowth rates of the rodent popu-Iations. Thus, predators or parasitesmay either affect a 'doomed rodentsurplus', which in any case woulddisappear, or drive the dynamics ofthe rodents.

Theoretical explanations and predictionsIntrinsic factors

Although great fluctuations inrodent populations were detectedmany years ago, it was not until the

L-

1950s that general explanations forthese fluctuations were proposed.The first to emerge involvedphysiological stress, thought todevelop at high density due toboth physical and social stimuli2.Later, intraspecifit aggression wasemphasizedi rodents were thoughtto experience benign and malignconditions alternatively, the latteraffecting both reproduction andsurvival for the benefit of thepopulation.

Such views were at odds with thenotion of individual selection, but itwas also difficult to imaginepopulations of the same qualityeither increasing or decreasingfrom the same density. Thus, a newtheory - that individuals are ofdifferent genetic quality in low- andhigh-density populations (the Chit-ty Hypothesis3) - was developed.Timid, well-reproducing animalsshould be selected against in in-creasing populations, and largepoorly reproducing and aggressiveindividuals should dominate athigh de-nsities. The increasingproportion of aggressive and poorlyreproducing individuals shouldlead to a population decline. Thispolymorphic system was - like thestress-regulated system - thoughtto produce very regular cyclesunder most environmental con-ditions.

Populations regulated by thesetwo intrinsic mechanisms shouldchange in size according to astereotypic pattem without anygeographic variation. Species withdistinct adaptations to food orclimatic factors should show inde-pendent cycles, and no commonpattern should emerge for differentrodent species occupying the samehabitat. Populations of predatorsmight fIuctuate in synchrony withrodents but they were not assumedto affect rodent dynamics, and nocommunity-wide patterns werepredicted.

TREE vol. 3, no. 8, August 1988

~

or biomass1o. The variability pre-dicted in predator effects is consis-tent with non-cyclic populationsbeing regulated by generalist pred-ators and cyclicity occurring only incommunities with few generalists.

Different types of parasites showdifferent capacities in regulatingtheir hast populationsl'. Micropara-sites (viruses, bacteria, etc.) of tencause epidemics and high mortalityin high-density host populations,and therefore will be potentialcauses of hast population crashes.Macroparasites (helminths, arthro-pods) have Ion g generation timesand ofte n less virulence, whichmakes them less probable agents

Fig. I. Weasels, here the least weasel (Muste/a niva/is), may play an important role in regular vole of rapid hast population decline.cycles. Photograph by Asko Kaikusa/o. Many. microparasites are rather

host-specific and might wellaccount for declines of single

quality between summer and main idea in small rodent research populations; high density, and im-winter, then large overwintering has been that predation will paired nutrition and immunity,populations leading to cyclic peaks deepen and prolong low-density may be necessary prerequisites.would not develop. However, phases caused by other factors9. It However, if pathogens are to causethe seasonal dynamics predicted has also been suggested that more synchronous cycles at the commun-for such conditions may change generalized predators switch to ity level, generalist microparasites(for instance to very low summer alternative prey in rodent declines have to be responsible. In such adensities) if other community and thus keep the main and case, high density of one hast couldprocesses (e.g. predation) are more alternative prey populations be sufficient to cause pronouncedimportant than food supply. covarying with some delay4. This declines in several species.

Predation has attracted less should occur especially in relative-attention than food as a factor ly simple communities with few --- o-. . influencing rodent fluctuations. alternative prey species. At the Simple predator-prey models pre- other end of the scale, very gener- I

dict regular cycles due to a time lag alized predators can stabilize prey ~in the numerical response of spe- populations by switching between Icialist predators (Fig. I), but the prey species according to density I~

J

General trends in den~itv variatinn~Rodents

Recent analyses (Box.l) haveshown that on ly a minority of micro-tine populations is cyclicI3.14.16Northem European CJethrionomysand Microtus species showa de-clining amplitude in their fIuctu-ations from the north towards thesouth and west (Table I). Northempopu!ations have a three to fauT (invery north em areas sometimesfive) year cyc!e, while southernScandinavian and British popu-!ations are usua!ly non-cyclic (Fig.3). Central Eurasian populations, atleast of CJethrionomys, are mostlynon-cyclic. }apanese populations ofC. rufocanus conform, by and large,to this non-cyclic pattem 17.

In contrast to Europe, NorthAmerican CJethrionomys and Mic-/ rotus species do not show anygeographical trends in cyclicityl4.16,though same American Microtuspopulations are distinctly cyclic.

I I I I I I I I Temperate American Microtus1 2 3 4 5 6 7 8 9 year populations ofte n seem to change

F' 2 G ,. d . f ,. ( b ) d l. (b l ) .. l . fl . between 'annual and multi-annua!'Ig. , enera Ize plcture o cyc IC a ove an non-cyc IC e ow mlcrotlne pop u atlon uctuatlons. . 1819 .,.Examples of cyclic populations can be found in Refs 32 and 33 and of non-cyclic ones in Refs 7. 16. 19 dynamlcs . ,resembllng In thls re-3nd 37. spect populations in a transition

\ \

196

TREE vol. 3, no. 8, August 1988

zone from cyclic to non-cyclicpopulations in central Scandinavia.Certain northem North AmericanCJethrionomys populations mayhave a 10-11 year cycle. CJethri-onomys rutiJus, for example, whichshows a four to five year cycle innorthern Scandinavia and hasshort-terin non-cyclic dynamics incentral Alaska and in the CanadianNorth-West Territories, has shownan II year cycle in Yukon (5. Gilbert1987, PhD thesis, University ofLund, Sweden). Lemmings - theclassical example of cyclicity - infact sometimes show more irregularfluctuations than many vaIepopulations2o.21.

The highest indices of cyclicityfor voles (Fig. 4) are reportedfrom northem Scandinavian popu-lationsl6, where the dynamics ofdifferent rodent species are strong-ly synchronized; this synchrony isgradually lost with decreasing cyc-licity southwardsl5. In regionswhere populations are general1ynon-cyclic, certain populations maynevertheless show regular andhigh-amplitude cycles, Such popu-Iations seem- to be associated withextensive areas of uniform habitats,e.g. afforestations in Britain andagricultural land in central'Europe 14.

The muskrat (Ondatra zibethica),. which is an unusual1y large micro-tine, demonstrates lO-li year cyc.iicity in North America22, whlIe in-troduced populations in northemScandinavia instead showa threeto four year cycle23.

The snowshoe hare (Lepus amer-icanus) has a lO-Il year cycle in themajor parts of its North Americanrange22. H oweve r, this cycle dis-appears at the southem distri-bution border in Wisconsin, Colora-do and Utah24. Mountain hares (L.timidus) in central parts of theUSSR are also claimed to haveapproximately a 10 year cycle, asdo jackrabbits (L. caJifomicus) incentral parts of the USA25, NorthemScandinavian mountain hares showthree to four year cycles26 on a verylow density level (except onislands) and cannot be consideredas primary cyclic populations (seebelow).

PredatorsThe cycle lengths of predators

are generally three to four years in

1!!7

northwestern Europel2 and 10-11years in North America22. However,recent studies demonstrate achange in predator patterns withlatitude in Scandinavia26, with thecycles disappearing towards thesouth. This applies to the red fax(VuJpes vuJpes) and to its alterna-tive prey, hares and grouse; therecruitment of alternative prey isdiminished during yale crashyears26. In southern Finland andSweden neither voles, predatorsnor alternative prey show regularcycles (Fig. 5).

ParasitesMost of the few studies on micro-

parasites in microtine rodentscome from zoonotic diseases,rodents being the main hast. Sever-al disease outbreaks have beenassociated with vole peaks in Scan-dinavia. Tularemia cases in humansshowa peak at and just after smallrodent peaks27, and the zoonoticdisease Nephropathia epidemica, akind of hemorrhagic fever, closelytracks CJethrionomys gJareoJuscycles28.29. Furthermore, the inci-dence of Nephropathia disappearssouthwards, at around ('DoN, wherethe cycles in this rodent speciesalso become weak or non-existentl4: The prevalence of viralantigens or antibodies has beenexamined in same microtinepopulations28-31 and there is sameevidente of a coincidence of viralepidemics and rodent declines.Certain viruses have been foundin several sympatric rodentspecies30.31. Very little is knownabout the joint or synergistic effectsof different pathogens in wildpopulations, but such synergism isknown to occur in laboratory ro-dents.

Observed patterns in relation to various

predictionsRegular small rodent cycles are

restricted to certain regions or,elsewhere, just to a limited seriesof years; the global pattern pre-dicted by theories of intrinsicmechanisms simply does not occur.This does not exclude the exist-ence of some inherent sodal orgenetic regulating mechanism thatis overridden by extrinsic factors32,but such a scenario is not en-

TREE vol. 3, no. 8, August 7988

Fig. 3. Generalized and partly tentative picture of thedistribution of strongly cyclic (dark), weakly cyclic'semi-cyclic' popu!ations (hatched) and more orless non-cyclic (light area) microtine populationsin northwestern Europe. From Ref. 16.

feeding strategy, food supply andsodal behaviour (territoriality ornon-territoriality in breedingfemales - see Fig. 4) may explainwhy folivores, for instance, havehigher breeding densities andgreater tendencies to cycles or out-breaks than granivores. On theother hand, distinct synchrony inprocesses such as the timing of thedecline is difficult to understandfrom nutritional factors, especially ifthe decline is not related to den-sit y, feeding strategy or social be-haviour of each species32. Temporalvariation in the production of dif-ferent plant material is obvl'Ouslynot 50 synchronized that it couldcause the type of rodent synchronyand crash found. for example, innorthem Scandinavia33, Floweringcycles seem to be mainly causedby rødent grazing33. A strict syn-chrony in a rodent community isalso hard to reconcile with plantdefence hypotheses unless de-fence communication between verydifferent food plants or plant pro-ducts is assumed (but seecriticism34 against 'talking trees').Recent studies do not support thefole of defensive chemicals in ro-dent cycles35.36. Neither do nutrientdynamics in the region of pro-nounced cycles in northem Scandi-navia explain microtine dynamics35.However, conditions in permafrostareas, for which nutrient dynamicshave been proposed as an explana-tjon of rodent dynamics, might be

ær ()~~

"'-v-

r-"'""

visioned in the main theories onintrinsic density regulation. Soda!behaviour may be involved in de-termin ing peak numbers (Fig. 4),but much confusion has resultedfrom the widespread beljet thatsuch factors are also causing thecyclic performance.

Interactions witli resourcesThe non-synchronous and non-

cyclic dynamics of rodents withdifferent feeding habits can partlybe understood as their indepen-

(b)

O)

.cOl.-10 Oc: .

C-~-oo

Q, 5.0C-

O)C1>

o>

1970 1972 1974 1976 1978 1970 1972 1974 1976 1978

Fig. 4. Density indices of the field vole, Mfcrotus agrestis, obtained with the same trapping method in (a) northem Finland (Pallasjårvj32) and (blsouthem Sweden (Revinge37). This species shows differences in peak numbers between cyclic and non-cyclic populations while, e.g. Clethrionomysspp. do not increase to very high numbers at cyclic peaks, and thus comply still better with the generalized picture in Fig. 2. Female territoriality in thepartly granivorous Clethrionomys spp. may limit reproduction and peak numbers while male M. agrestis, as many other pronounced folivores, arepolygynous and females non-territorial.

198

dent reactions to spedfic re-sources. Non-cyclic patterns inmany regions could be responsesto nutritional factors, possibly ininteraction with sodalones, and toclimatic hazards. The farer occur-rence of cyclic populations of anesped es in ane habitat and non-cyclic anes of another spedes inother habitats in the same regionmay be related to feeding habitsand the abundance of suitablehabitats.

A common relationship between

TREE vol. 3, no. 8, August 1988

,...

~,..z..., ",I

..

..

..

..

c.,

C7

differ:ent20. Experimental additionsof food generally cause temporaryincreases in density but they donot determine the lang-termdynamicsl6.

Interactions with predatorsObservations and experiments

on non-cyclic rodent populations insouthern Scandinavia indicatedthat most (80%) rodent individualswere taken by generalist predators(foxes, buzzards, cats, etc.) withabundant alternative prey37. Spec-ialist predators (mustelids, certainowls, etc.) may be more importantin limiting northern cyclic rodentpopulations, and tjeld data fromnorthern Finland indicate indirectlythat they have a depressive func-tjon during rodent declines32.

Also, the decline in the degreeof interspecific synchrony amongmicrotine rodents from northern tosouthern Scandinavia 15 suggests achange in the main limiting factors,possibly from specialist predationin the north to generalist predationon folivores and food limitation ofgranivores in the south. For gener-alist predators, snow diminishesboth rodent availability and num-ber of alternative prey; specialistweasels are well suited for hun tingucder snow, where they are alsoprotected from their own predators.Rodent cyclicity shows a significantpositive correlation with snowcover in Scandinavia.

At the southern distribution limitof the snowshoe hare, 'facultative'(generalist) predators have beenimplicated as the limiting factor38,causing non-cyclic dynamics. Whena north Scandinavian fax popu-Iation (being rather specialized onsmall rodents in this region) wasseverely reduced by a sarcopticmange outbreak, mountain haresincreased to much higher densitylevels than befare and becamenon-cyclic39.

Interactions with pathogens and parasitesOutbreaks of opportunistic

pathogens, commonly found in theenvironment, can cause episodicdeclines in microtine popu-lations31, and this may contribute tothe unpredictability of microtinedynamics in many regions. Latentinfections may also be common,and may become acute when im-

199

munological function is impaireddue to nutritional or stressfactorsI8,21,3o. The occurrence of cer-tain murine viruses in severalmicrotine species suggests that thedisease may be important also at acommunity level. Generalist micro-parasites could be an alternative tospecialist predators in causing syn-chronous dynamics.

ConclusionsWhen populations of the same

species show clearly distinct dy-namics in different regionsl4, differ-ences in extrinsic factors areimplied. Geographical (or habitat)trends may indicate clinal variationin the effects of different factors.Instead of considering vague multi-factorial hypotheses, alternativecausal mechanisms for furtherstudy can be suggested, as indi-catedabove.

New data and re-analysis of olddata demonstrate three changes inaur understanding of rodent dy-namics: (I) Most rodent pop u-Iations are not cyclic; regu.Iar cyclesare restricted to certain areas, whileirregular outbreaks may occurworldwide; (2) Many other com-ponents of the surrounding com-munity which interact directly orindirectly with rodents show syn-chronous population changes; (3)The earlier weak emphasis on foodor nutrition as a driving factor ofcycles is changing to predators andprobably also disease.

The widespread idea that allrodent fluctuations folIow the samepattern and are generated by thesame intrinsic mechanism has to beabandoned. However, no single ex-trinsic factor may emerge to replacethe old theoryj instead, various fac.tors may play important roles indifferent geographic locations, andpossibly in the same locations atdifferent times. The composition ofthe local biotic community, and itsinteraction with physical factors,such as snow cover and habitatheterogeneity, may be decisive.The data indil:ate that many rodentpopulations have only seasonalfluctuations but that habitat and/orcommunity changes, either tempor-arr or permanent, may influencesuch a regulation and lead to out-breaks or cycles with their own typeof regulatory processes.

Fig. 5. Regions with (shaded) and without (light) gamecycles in Scandinavia. Reproduced, with permission.from Ref. 28.

AcknowledgementsWe are grateful to Per Angelstam. IIkka

Hanski and Stuart L. Pimm for comments.

ReferencesI Krebs, (.J.and Myers,I.H. (1974) Adv. Ecol.Res. 8, 267-3992 Christian, 1.1. 11950) j. Mammal. 31,249-2593 Chitty. D. I 1967) Proc. Ecol. Soc. Aust. 2,51-784 Lack, D. 11954) The Natural Regulation ofAnimal Numbers, Clarendon Press5 Kalela, O. (1962) Ann. Acad. Sci. Fenn. Ser. .

A466.1-386 Pitelka, F.A. (1964) in Grazing in TelTestrialand Marine Environments (Crisp. D.I., ed.),pp. 55-56, Blackwell7 Negus, N.C.. Berger, P.I. and Brown. B.W.11986) Can. j. Zool. 64, 785-7928 Haukioja,E.(1980)Oikos35.202-2139 Pearson, O.P. (1966) j. Anim. Ecol. 35.217-23310 Murdoch,W.W.and Oaten,A. (1975) Adv.Ecol. Res. 9,1-131I1 Anderson, R.M. and May, R.M. (19791Nature 280,361-36712 Finerty, I.P. (1980) The PopulationEcology of Cycles in Small MammaIs:Mathematical Theory and Biological Fact,Yale University Press13 Henttonen, H.. McGuire, A.D. andHansson, L. 11985} Ann. Zool. Fenn. 22,221-22714 Hansson, L. and Henttonen, H. (1985)Oecologia 67. 394-40215 Henttonen. H. and Hansson, L. Holarct.Ecol. I1 (in press)16 Taitt, M.I. and Krebs. C.J. (1985) in Biologyof New World Microtus (Tamarin, R.H.. ed.J,pp. 567-620, The American Society of

Mammalogists17 Saitoh. T. (1987) Oecologia 73, 382-388i8 Mihok. S.. Tumer, B.N. and Iversen, S.L.(1985) Ecol. Monogr. 55, 399-42019 Getz, L.L.. Hofman, I.E.. Klatt. B.I.. Cole.F.R. and Lindroth. R.L. (1987) Can. j. Zool. 65. ,'O-:i-'1.1317-1325 c~,'

TREE vol. 3, no. 8, August 1988

34 Fowler,S.V.andLawton,J.V.(1985)Am.Nat. 126, 181-195

35 Jonasson, S., Bryant, J.P., Chapin Ill, F.S.and Andersson, M. (1986) Am. Nat. 128,394-40836 Lindroth, R.L. and Batzli, G.O. (1986)I. Anim. Eco/. 55, 431-449

37 Erlinge, S.. Goransson, G.. Hansson, L..Hogstedt, G.. Liberg, O:, Nilsson, I.. Nilsson,T.. von Schantz, T. and Sylven, M. (1983)Oikos 40, 36-52

38 Sievert, P.R. and Keith, L.B. (1985) I. Wildl.Manage. 49, 854-865

39 DaneIl, K. and Homfeldt, B. ( 1987)Oeco/ogia 73, 533-536

27 Homfeldt. B. (19781 Oecologia (Berlin) 32.141-15228 Brummer-Korvenkontio. M.. Henttonen.H. and Vaheri. A. (.I 982) Scand.l. /nfect. Dis.(Suppl. 36).88-9129 Niklasson. B. and LeDuc. J. (1987) J./nfect.Dis. 155. 269-27630 Kaplan. C.. Healing. T.D.. Evans. N.,Healing, L. and Prior, A. (1980) ,. Hyg. 84,285-29431 Descoteaux. J.P.and Mihok.S. (1986)J. Wildl. Dis. 22. 314-31932 Henttonen, H.. Oksanen. T.. Jortikka. A.and Haukisalmi, V. (1987) Oikos 50,353-36533 Andersson. M. and Jonasson. S. (1986)Oikos 46, 93-106

20 Batzli. G.O.. White. R.G.. MacLean. Ir. S.F..Pitelka. F.A. and Collier. B.D. (1980) in AnArctic Ecosystem: the Coastal Tundra atBarrow, Alaska (Brown. I., Miller. P.C..Tieszen. L.L. and Bunneil. F.L., edsl, pp.335-410. Dowden, Hutchinson and Ross21 Laine, K. and Henttonen. H. (1983) Oikos40.407-41822 Bulmer. M.G. (19741 f. Anim. Ecol. 43.701-71823 Danel!. K. ( 1985} Acta Theriol. 30. 21 ~22724 Buehler. D.A. and Keith. L.B. (1982} Can.

FieldNat.96.1~2925 Keith. L.B. ( 1983) Oikos. 40. 385-39526 Angelstam. p., Lindstrom. E. and Widen, P.(19851 Holarct. Ecol. 8.285-298

Michael J. Hutchings

A/thougn ec%gists nave spent mucn effortin ana/ysing the foraging behaviour ofanima/s, tne study of p/ants as foragingorganisms is a re/ative/y unexp/ored sub-;ect. Tnere is often, however, much greaterpotentia/ for ana/ysis of foraging benaviourin p/ants tnan in anima/s. On/ille mostanima/s, many p/ant species /eave perma-nent or semi-permanent records of tneirforaging activities because tneir resource-acquiring structures (primari/y /eaves androots), persist for a considerab/e time, asa/so do tne structures (trunks, branches,st%ns, runners or rhizomes) which enab/e/eaves or roots to be pro;ected info particu-lar positions in the habitat. In addition,p/ant ec%gists are not burde ned with tnedifficu/ties associated witn determining howchanges in foraging benaviour affect fftnessin anima/s', because p/ant mass (or, in tnecase of dona/ species, number of rametsproduced) , is usua/ly dose/y corre/atedwith fftness.

The term 'foraging' can be de-flned as the process whereby anorganism searches or ramifies with-in its habitat in the activity of at-quiring essential resources2. Unlessthe pattern of search changes whenthe organism encounters patches ofhabitat containing different concen-trations of resources. however. ac-quisition of resources will not beachieved efficiently. In plants.changing search patterns areusual-

I

II

IIII

III

Michael Hutchings is at the School of BiologicalSciences, University of Sussex, Falmer, Brighton,Sussex BNI 90G, UK.

@ 1988. Elsevier Publications. Cambridge 0189--5347/88/$0200200

ly achieved through morphologicalplasticity2-6.

Interpretations of experimentaldata on plant form in terms of dif.ferential foraging were almostentirely lacking from the ecologicalliterature until recently. For suchinterpretations to be possibleexperiments on plants in controlledenvironments, rather than obser.vations of plant behaviour undertjeld conditions, are essential. Thisis because the natural environmentis highly heterogeneous even at avery small scalei the genetical ori-gin (and therefore the morphologi-cal properties) of plants in the tjeldmay be uncertain even 'within aspecies, and the effects on mor-phology of the interactionsbetween the plants under studyand neighbouring plants are verycomplex7. Altogether these con.siderations confound attempts tointerpret morphological differenceswithin a species from place to placein the tjeld in terms of differentialforaging.

Greenhouse experiments on foraging

Most references to plants asforaging organisms con cern her-baceous clonal species that growprimarily in the horizontal plane,branching and spreading laterallyrather than acquiring height8.-lo.Most of the available data aboutplant foraging are from such spe-:ies. The behaviour of these spe-:ies is clearly easier to analyse than

acquisition is primarily carried outby two-leaved structures termedramets, which are produced atevery node along stolons whichcreep over the soil surface (Fig. I J.The plant also produces roots at itsnodes. Two furtheistolons can growfrom lateral buds at each node, 50that the clone develops as a bran-ched, connected population oframets. For descriptive purposes,the principal stolons and theramets they bear are described asprimaries, whereas lateral budsgive rise to secondary (and tertiaryJstolons and ramets. Although theramets are strictly determinate inform, the whole done is markedlyplastic in structure.

Plants of G. hederacea for experi-mentation can be proliferated froma single clone, 50 that allobservedvariation in morphology can beascribed to differences in the grow-ing conditions applied to replicateclones. Each ramet is large enough,and far enough from its neighbourson the same stolon, to be sub-jected in experiments to its ownpersonalized set of growing con-ditions. Thus it could be provided,if wished, with a supply of light,nutrients and water which is differ-ent from that given to every otherramet. Experiments have been per-formed to compare the morphologyof whole clones given either ampleor very limited supplies of eithernutrients or light2.5.7.

In addition to the expectedreduction in biomass of doneswhen resource supply was low,there were marked differences inthe morphologies of clones receiv-